AI Prompts for Brainstorming Sessions

Research on group ideation consistently shows that the average brainstorming session produces fewer unique ideas per person than the same group would generate thinking alone. The culprit is social dynamics: people anchor on early ideas, defer to the most senior voice, and self-censor to avoid looking foolish. The problem isn't the people. It's the format.

These ai prompts for brainstorming restructure the session format itself — not just what you ask, but the sequence that produces better output.

Quick-Start (Copy This Right Now)

If you need to run a brainstorm in the next hour and want a better structure than "everyone shout ideas":

I'm facilitating a brainstorm on [topic] with [number] people from [describe roles/backgrounds]. The goal is to generate [what you want to produce — solutions, ideas, campaign concepts, etc.].

Design a 45-minute brainstorm structure that:
1. Starts with individual silent ideation before any group sharing
2. Uses a structured prompt sequence that builds from divergent to convergent
3. Includes one technique that surfaces ideas from quieter participants
4. Ends with a clear output we can act on immediately after the session

What this does: Produces a facilitation plan, not just a list of questions. The "individual silent ideation first" instruction is the single most evidence-backed change you can make to a brainstorm — it eliminates anchoring and increases output diversity by 40–60% according to most creativity research.

⚡ Pro tip: Print or share the AI-generated session structure with participants before the session, not during it. When people know the sequence in advance, they do pre-thinking that makes the live session substantially more productive.

Understanding the Variables

[Topic] — Be as specific as possible. "How to improve customer retention" is a topic. "How to re-engage customers who stopped using the product after the first 30 days" is a brainstorm prompt. Specificity dramatically narrows the solution space in a useful way.

[Roles/backgrounds] — A group of engineers brainstorms differently than a mixed group of engineers, marketers, and salespeople. The AI will design different facilitation techniques based on group composition.

[Output type] — Ideas you'll rank? Solutions you'll prototype? A concept you'll pitch? The output format shapes the session structure. A brainstorm that ends with "here are 40 ideas" is different from one that ends with "here are 3 concepts ready to prototype."

Step-by-Step: Running an AI-Assisted Brainstorm

Step 1: Pre-Session Idea Generation

Send this to participants 24 hours before:

We're brainstorming about [topic] tomorrow. To prepare, spend 10 minutes and write down: (1) the most obvious solution everyone will suggest, (2) the solution you'd personally find most interesting to try, (3) one approach that might seem too extreme but could actually work, (4) what you think we're not even asking about that we should be.

What this does: Primes individual thinking before group dynamics kick in. The "too extreme" and "what we're not asking" categories consistently produce the most novel ideas — because they force thinking outside the implicit constraints of the stated problem.

Step 2: The Divergence Prompt

Early in the session, before any evaluation:

For the problem [state problem], generate 20 ideas with no filtering for practicality. Include: 5 that are immediately actionable, 5 that would require significant resources, 5 that are unconventional, and 5 that address a different version of the problem than the one stated.

What this does: Forces quantity and variety before quality. The "5 that address a different version of the problem" category is where the real innovations often hide — they reframe the problem rather than solving it as given.

⚡ Pro tip: Run this prompt yourself before the session and bring 5–6 seeded ideas to share after participants have shared their own. Seeded ideas from a facilitator planted before participants have spoken anchor everyone to those ideas. Seeded ideas shared after participants have gone first simply expand the pool.

Step 3: The Convergence Prompt

After ideas are on the table:

Here are the ideas generated in our brainstorm: [paste list]. Help us converge by: (1) grouping similar ideas into 5–7 themes, (2) identifying the 3 ideas with the highest potential for impact, (3) identifying the 1 idea that most challenges our current approach, (4) suggesting which two ideas might be stronger combined than alone.

What this does: Applies a structured lens to a messy list — the output of every brainstorm that's working. Without this convergence step, most sessions end with a long list and no decision about what to do with it.

⚠️ Common mistake: Skipping the convergence step because "we want to preserve all the ideas." You're not throwing ideas away by converging — you're deciding what to prioritize first. The rest of the list stays available. But pretending all 40 ideas are equally worth pursuing is how brainstorms produce no action.

Step 4: The Decision Output

Based on our brainstorm, we want to leave this session with a specific next step. Here are our top 3 ideas: [list]. For each one: (1) what would a 2-week experiment to test this look like, (2) what's the cheapest way to get signal on whether this will work, (3) who needs to be involved to move this forward?

What this does: Converts brainstorm output into a testable plan. This is the step that separates sessions people remember as productive from sessions people remember as fun but vague.

Pro-Level Variations

For technical brainstorms:

We're brainstorming technical approaches to [problem]. Include in the idea generation: one approach that optimizes for speed, one that optimizes for scalability, one that optimizes for maintainability, and one that prioritizes user experience over technical elegance.

For constraint-based brainstorms:

We're brainstorming with these hard constraints: [list]. Generate ideas that work within the constraints, then generate ideas that challenge the constraints themselves. Which constraints are actually soft — negotiable if the idea were good enough?

For async brainstorming:

I want to run a brainstorm asynchronously across my team over 48 hours. Design a prompt sequence they can respond to on their own time that: prevents people from being anchored by others' responses, still builds toward a converged output, and doesn't require a synchronous session to produce an actionable result.

Troubleshooting Common Issues

"Everyone keeps suggesting the same ideas." Use the divergence prompt's "5 that address a different version of the problem" category before the group shares. Forcing a reframe breaks the anchoring.

"The session ends without a decision." Build the Decision Output prompt into your facilitation plan as the explicit last step. If there's no time, that's a structure problem — not a decision problem.

"Senior voices dominate." Run Step 1 (pre-session ideation) as anonymous written input first. Facilitators collect and display responses without attributing them. When ideas are detached from seniority, evaluation becomes more honest.

Your Turn

Take your next scheduled brainstorm and run the Quick-Start prompt to build a proper structure before the session. Share the structure with your team 24 hours in advance. Notice how the session quality changes.

Save your brainstorming prompt templates in PromptABCD — particularly the divergence and convergence prompts, which are applicable to virtually any brainstorm topic.

The Solo Brainstorm: When You Need Ideas Alone

Not every brainstorm involves a team. Solo ideation sessions have their own failure mode: you circle the same 3–4 ideas and can't break out of your own frame.

I'm brainstorming alone about [topic]. I've already thought of these ideas: [list your current ideas]. My constraint is: [what you're working within].

Help me break my frame: (1) What are 5 ideas that start from a completely different assumption than the ones I've already thought of? (2) What would a person with a background in [different field — e.g., "game design," "behavioral economics," "military strategy"] suggest? (3) What would the most contrarian answer be?

What this does: Breaks the self-referential loop that solo brainstorming often creates. The "different field" parameter is the most productive — it imports mental models from outside your normal frame and applies them to your specific problem.

⚡ Pro tip: When doing solo brainstorming with AI, push past your first batch of outputs. The first 10–15 ideas are usually variations on obvious approaches. Run the prompt a second time with: "None of the above were surprising. Generate 10 more that are weirder and less safe." The second batch is where genuinely novel ideas appear.

After the Brainstorm: What to Do With 40 Ideas

Most brainstorm outputs die in the document they were created in. Here's the prompt that converts them into action:

Here are all the ideas from our brainstorm: [paste list]. We need to choose one to run a small experiment on within the next two weeks. Select the one that: (1) has the highest chance of teaching us something regardless of outcome, (2) requires the fewest resources to test, (3) would not prevent us from trying other ideas if it doesn't work.

What this does: Applies a "learning value" lens instead of a "best idea" lens to the selection. The best idea is often the hardest to test. The one that teaches you the most, fastest, is often the right starting point. Save this post-brainstorm prompt in PromptABCD alongside your session structure template — together, they cover the full brainstorming cycle.

⚡ Pro tip: After every brainstorm session, regardless of outcome, run a one-minute retrospective: "What made this session more or less productive than our last one?" Over 6 months, you'll build a picture of exactly which session formats, prompt sequences, and facilitation moves produce the best results for your specific team — no productivity book required.